Spatiotemporal variability of streamflow and attribution in the Three-Rivers Headwater Region, northwest China

Global change has become an indisputable fact (IPCC 2007). As an important aspect of this change, climate warming has already had an important influence on natural ecosystems and water resources (Houghton et al. 2001). Climate warming can accelerate the water cycle and alter the spatial and temporal distribution patterns of the regional water resources (Boer et al. 2000; Nijssen et al. 2001; Labat et al. 2004; Ramanathan et al. 2005). There has been an increase in studies into the effects of a changing climate on hydrology and water resources over the past decades (e.g. Lettenmaier et al. 1999; Xu & Singh 2004; Chen et al. 2006; Gardner 2009; Zhang et al. 2009). In China, many scholars have conducted systematic and in-depth research into the spatial and temporal distribution patterns, evolution, and factors influencing changes in the water resources of the Yellow River Basin (YRB), Yangtze River Basin (YARB), Haihe River Basin, Northwest Inland River Basin, and other basins (Fu et al. 2004; Hao et al. 2008; Yang & Tian 2009; Shi et al. 2013). Some researchers have focused on water cycle-related factors, including changes to and interrelationships between runoff, precipitation, evaporation, groundwater, and soil water, and have been especially concerned with the historical evolution and possible future change to water resources under various different climate scenarios and changing environments, including human-induced land use and land cover changes (Hao et al. 2008; Yang & Tian 2009; Zhang et al. 2009; Shi et al. 2013). Research results have shown that the runoff in some Chinese river basins, especially those in the north of the country, has been decreasing to varying degrees, which has reached a significant level in some parts of the basins (Liu & Zheng 2004; Hao et al. 2008; Wang et al. 2009; Yang & Tian 2009; Zhang et al. 2009; Ma et al. 2010; Yang et al. 2010; Shi et al. 2013). Additionally, the results have revealed some of the causes of runoff variation, among which climate warming and increased evaporation, volatility of precipitation, increased water consumption for industrial and agricultural production, and underlying surface changes caused by human activities have been considered as the main causes for the sharp drop in runoff (Liu & Zheng 2004; Hao et al. 2008; Wang et al. 2009; Yang & Tian 2009; Ma et al. 2010; Yang et al. 2010).

The research carried out to date has some shortcomings and deficiencies, however. Firstly, previous studies have focused more on identifying changes and performing attribution analysis for the water volumes of major rivers on national and regional scales, but little attention has yet been given to small-scale basins. Some specific regions have insufficient geographical areas to be considered in water resources planning, eco-environmental construction, and other local governmental policies. Secondly, a number of these previous studies have not adequately revealed the intra-annual and inter-annual change of runoff, leading to single and one-sided attribution analysis of runoff variation, which is not conducive to a comprehensive understanding of the evolution of water resources. China covers a vast area of land, with large geographical differences between the north and south of the country. As a result, the water cycles of different basins have markedly different characteristics, and even the features and problems faced by different sections of the same basin. In general, in the eastern monsoon regions, precipitation recharge is the major source of runoff, accounting for more than 80% of the total, and the variations in runoff are basically correlated with monsoon strength. In contrast, in the Northwest Inland River Basins, precipitation recharge is not the only source of runoff, as meltwater from glacier and snow accounts for more than 50% of the total runoff in some basins. In these basins, the intra-annual distribution of runoff is affected by the monsoon, but it is also connected to the amount of spring snowmelt and other factors (Xu et al. 2004; Chen et al. 2006; Hao et al. 2008).

The Three-Rivers Headwater Region (TRHR) in China plays an important role in East Asian and global river systems (Liu et al. 2008a, 2008b; Fang 2012; Tong et al. 2014). The TRHR has an altitude ranging between 3,335 and 6,564 m, with glaciers, snow cover, permafrost and wetland widely distributed across the area, making its river systems unique, and this has been the key point in studies into cryosphere hydrology and the water resources of the Tibetan Plateau (Liu et al. 2008a, 2008b; Fang 2012; Tong et al. 2014). Under the background of a changing climate, this area has experienced glacier recession, snow melting, permafrost degradation, amongst other warming phenomena, all of which constitute a serious threat to water conservation (Liu et al. 2008a, 2008b; Tong et al. 2014). Previous studies into climate change and water resources in the TRHR have typically used a small number of basins to study the long-term trends in runoff for the three basins, but have thus far paid little attention to the area's sub-basins (Bing et al. 2011; Zhang et al. 2012).

Therefore, the present study makes use of observational data from ten hydrological stations and nearly 20 meteorological stations across the TRHR to pursue the following objectives: (1) to investigate long-term streamflow and climate change in the major rivers of this region, focusing on the analysis of intra-annual and inter-annual change patterns; and (2) to determine the change in streamflow and its relation to precipitation, evapotranspiration, glacier melting, and human water consumption. Furthermore, we try to give the contribution ratio of climate variability, glacier melting, and human water consumption to streamflow. The primary goal of this study is to evaluate the impact of climate change and human activities on streamflow, and to provide basic information for modelling, the forecast of water resources, and in local governmental planning and decision-making with regard to water resources.

The TRHR is located in the hinterland of the Tibetan Plateau, in southern Qinghai Province, and contains the headwaters of the Yellow River, Yangtze River, and Lancang River. Geographically, it is located between 31 °29′ N and 36 °12′ N, and between 89 °45′ E and 102 °23′ E. It has a total area of approximately 39.6 × 10⁴ km² (Liu et al. 2008a, 2008b; Fang 2012; Zhang et al. 2012). The topography of the TRHR is mainly mountainous, with an altitude ranging from 3,335 to 6,564 m, and high mountains with altitudes of 4,000–5,800 m forming the main skeleton of the topography. The climate of the TRHR is a typical plateau continental climate, with alternating hot and cold seasons, distinct wet and dry seasons, a low annual temperature, long sunshine hours, and intense solar radiation (Liu et al. 2014). The spatial distributions of soils show significant vertical zonation, with, from high to low altitude: alpine cold desert soil, alpine meadow soil, gray, brown soil, chestnut soil and mountain forest soil (Liu et al. 2008a, 2008b; Shao et al. 2009; Fan et al. 2010).

Monthly streamflow records were collected for the same period from ten major river gauge stations (Table 1) across the TRHR. These gauge stations include six stations on the mainstreams of the YRB, YARB, and Lancang River Basin (LRB), and four stations on the tributaries of these three rivers. In instances when runoff data were missing, we used runoff data from similar rainfall conditions at other times as a replacement.

The digital elevation model (DEM) with 90 m resolution derived from the US Geological Survey (www.usgs.gov/), and the glacier area data in the TRHR was taken directly from the Cold and Arid Regions Science Data Center (http://westdc.westgis.ac.cn/). The land use maps in TRHR in 1985, 2000, and 2010 were derived from the Cold and Arid Regions Science Data Center.

Flow duration curves (FDCs) were constructed for daily, monthly, and annual river discharges, over each time interval of interest. Each value of discharge (Q) has a corresponding exceedance probability (P), which indicates the percentage of time when a given flow rate is equalled or is exceeded (Melesse et al. 2010). An FDC is, therefore, simply a plot of Q_p, the Pth quantile or percentile of streamflow, against the exceedance probability P (Smakhtin 1999). The FDCs in this study were used to determine streamflows with 5, 50, and 95% exceedance probabilities. The ratios of (Q₅/Q₅₀) and (Q₉₅/Q₅₀) were then used to examine changes in low and high flow patterns.

According to the published literature in China (Sun et al. 2006) and elsewhere (Zhang et al. 2001), the ω parameter values were assigned as 2.0 for high-cover woodland (where forest cover >30%), 1.0 for low-cover woodland (where forest cover <30%), 0.5 for grassland and cropland, 1.0 for shrubland, 0.1 for building and barren land. In this study, the coverage of land uses was known from the land-use maps; according to the land-use maps in 1985, 2000 and 2010, we obtained the mean value of different land-use coverages. Then, the sensitivity of streamflow to precipitation (β) and streamflow to potential evapotranspiration (γ) could be calculated by Equations (9) and (10), respectively. The change in streamflow caused by climate variability (ΔQ_clim) was calculated as Equation (8).

The annual average temperature for the entire region was 0.27 °C, with a maximum value in 2009 (1.76 °C) and a minimum value in 1957 (−1.00 °C) (Figure 2(b)). The temperatures thus display a clear increasing trend (0.31 °C/10 a), particularly over the 20 years since 1991. The rate of temperature increase during 1991–2012 (0.68 °C/10 a, P < 0.001) was significantly greater than that during 1956–1990 (0.18 °C/10 a, P < 0.01). Considering the spatial distribution of temperatures, the three basins may be ordered LRB (2.42 °C) > YRB (1.2 °C) > YARB (−2.5 °C), with the LRB showing the highest temperature. The order of rate of temperature increase in the three basins is YARB (0.40 °C/10 a) > LRB (0.37 °C/10 a) > YRB (0.31 °C/10 a), with the YARB showing the largest warming amplitude.

Based on our preliminary analysis of changes in precipitation and temperature, we found that the average annual temperature and annual precipitation in the TRHR has exhibited a significant increasing trend since 1990, with the rate of increase during 1991–2012 much larger than that during 1956–1990 (Table 2). This increase in temperature and precipitation may therefore be expected to exert some effects on streamflow in this region. As such, the discharge records were divided into two time periods, before 1990 and after 1990, in order to analyse the inter-annual and intra-annual variations in streamflow, by comparing the statistic coefficients from these two periods.

The average non-uniformity coefficient (C_y) of runoff in the TRHR was relatively large, with the value of both periods (1956–1990 and 1991–2012) being generally greater than 0.66 (Table 3). Comparing the two periods, with the exception of a small number of individual stations, C_y during 1991–2012 was generally less than that during 1956–1990. C_y in the YRB (Tangnaihai) decreased by 0.04, in the YARB (Zhimenda and Tuotuohe) it decreased by 0.01, and in the LRB (Xiangda) it decreased by 0.05. It should be noted that the total runoff in the TRHR increased during 1991–2012, but the C_y decreased during this time.

Table 4 shows that the annual runoff in the LRB (Xiangda and Xialaxiu) and the YARB (Zhimenda and Tuotuohe) have an overall increasing trend, however the variation of the main stream and branches of the YRB was inconsistent: the runoff in Tangnaihai, Maqu, and Tongren decreased slightly, but increased in Jimai and Huangheyan. Considering the seasonal variation, the runoff in the YRB (Tangnaihai) decreased most in the autumn, and also decreased in the spring, while it increased in the summer and winter. In the YARB (Zhimenda and Tuotuohe) and LRB (Xiangda and Xialaxiu), however, the runoff increased in all four seasons, with summer showing the greatest increase, and winter the smallest.

The FDCs of each station in the TRHR during the periods 1956–1990 and 1991–2012 are not shown, but the statistics are shown in Table 5. The high runoff and low runoff in the YRB both show varying degrees of decreasing trend. The trend in high runoff was between −11.1 and −29.7%, with the greatest decrease in Huangheyan, and the least decrease in Maqu. The variation in low runoff was between −6.0 and −99.7%, with the smallest decrease in Jimai, and the greatest decrease in Huangheyan. Taking the Tangnaihai station as an example, Q₅/Q₅₀ decreased by 5.0% between the 1956–1990 and 1991–2012 periods, but Q₉₅/Q₅₀ remained unchanged, indicating that the high flow of YARB decreased; for the YARB (Zhimenda station), Q₅/Q₅₀ and Q₉₅/Q₅₀ decreased by 10.1% and 7.1% between the two periods (1956–1990 and 1991–2012), respectively; in the LRB, the high runoff decreased while the low runoff increased. The amplitudes of variation of Q₅/Q₅₀ and Q₉₅/Q₅₀ at Xiangda station reached −12.8% and 32.4%, respectively.

In this study, the coverage of land use was known from the land-use maps, according to the land-use maps in 1985, 2000 and 2010, we obtained the mean value of different land-use coverages. Then sensitivity coefficients of runoff to changes in precipitation (β) and potential evapotranspiration (γ) could be, respectively, calculated as Equations (9) and (10). The result showed that β = 0.61 and γ = −0.42, revealing that the change in runoff was more sensitive to precipitation than to potential evapotranspiration. So the change in streamflow caused by climate variability (ΔQ_clim) was calculated as Equation (8).

The results indicated that in the TRHR, the proportional change in annual runoff due to climate variability accounted for >85% of the observed change, while anthropogenic activity and glacier melting was responsible for ∼15% (Table 7). The contribution rates of anthropogenic activity in the YRB and LRB were 7.7% and 6.8%, respectively, which were a little higher than that in YARB (5.1%) due to the higher population densities and greater anthropogenic activity. The contribution of glacier melting in the YARB and LRB were 9.5% and 6.8%, respectively, which were obviously higher than that in YRB (2.0%) due to the higher distribution densities of glaciers. During 1991–2012, the effects of climate variability and anthropogenic activity on runoff showed a significant difference. The glacier melting and human activities in the YRB made a negative contribution to runoff variation, however, they made a positive contribution in YARB and LRB. It can be inferred that the increase in runoff during the change period was mainly due to climate variability. Fortunately, adopting exclusion measures for preventing grassland degradation have increased vegetation recover to a certain extent, which has had a positive effect on runoff.

1. Precipitation and temperature in the TRHR exhibited significant increasing trends over the last 57 years, and the rate of increase during 1991–2012 was notably faster than that during 1956–1990. In the three basins, the rate of increase is ordered YARB > LRB > YRB.

2. The annual runoff of the LRB and YARB showed an increasing trend, while the runoff in the mainstream of the YRB was slightly reduced. The intra-annual distribution of runoff shifted gradually from a double peak pattern to a single peak pattern. In the YRB and YARB, both high and low flow together decreased or increased, respectively. In the LRB, the high flow decreased while the low flow increased.

3. In the TRHR, the proportional change in annual runoff due to climate variability accounted for >85% of the observed change, while anthropogenic activity and glacier melting was responsible for ∼15%. The contribution rates of anthropogenic activity in the YRB and LRB were a little higher than that in YARB due to the higher population densities and greater anthropogenic activity. The contribution of glacier melting in the YARB and LRB were obviously higher than that in YRB (2.0%) due to the higher distribution densities of glaciers.

This research was jointly funded by key consulting project of Chinese Academy of Engineering (2014-XZ-31) and Chinese Research Academy of Environmental Sciences special funding for basic scientific research (2014-YKY-003).